📋 README.md file for code accompanying the paper
This repository is the official implementation of paper [A. Pourafzal and A. Fereidunian, "A Complex Systems Approach To Feature Extraction For Chaotic Behavior Recognition," 2020 6th Iranian Conference on Signal Processing and Intelligent Systems (ICSPIS), Mashhad, Iran, 2020, pp. 1-6, doi: 10.1109/ICSPIS51611.2020.9349551].
This code is written in Jupyter NoteBook. To install the requirements in the corresponding environmnet, run this command in Anaconda Powershell Prompt:
conda install numpy matplotlib jupyter scipy sklearn
All the data could be generated using the following notebook,
\Generating Time-Series.ipynb
However as it is a time consuming process, final features are provided in
\Lorenz_System_Features.csv
📋 In the code, a propotion of 80 percent of this dataset is utilized as train data and the rest is for test.
Three different machines of SVM, KNN and Random Forest with tuned hyper-parameters are trained in the following notebook
\Classification.ipynb
The best model achieves the following performance as:
| Model name | Precision | Recall | F1 |
|---|---|---|---|
| SVM | 0.82 | 0.77 | 0.78 |
| Random Forest | .092 | 0.92 | 0.92 |
| KNN | 0.86 | 0.85 | 0.85 |
📋 Any modification, adaption and th use of this source code must be credited by citing the original paper